December 1997
Microsoft Corporation
One Microsoft Way
Redmond, Washington 98052, USA
Introduction
Updated in 2023 to include the original slides
The Internet represents a fundamental change in the landscape for consumer electronics.
Consumer devices have become increasingly intelligent as the capabilities of computation become available to even the simplest devices.
With the advent of the Internet they can now participate in a larger infrastructure. They can cooperate with each other and can take advantage of the infrastructure to enhance their capabilities. They can also allow themselves to be controlled by other devices. In particular, the personal computer can serve as a means of enhancing a device without burdening the user with the complexity of learning how to operate the device.
The key is to share the common, IP (Internet Protocol) infrastructure. Sharing the common infrastructure provides for economies of scale. More important, it provides for common protocols that allow independently designed devices to cooperate.
The Internet
There is no longer a need to explain what the Internet is -- it has become a global phenomenon. For many, though, the World Wide Web represents the Internet. But the Web is just one use of the Internet. It has been key to the recent explosive growth of the Internet.
The Web itself is just one application that "rides" the Internet. Tim Berners-Lee was able to create the Web by writing a simple application on his personal computer and sharing it with others. The URL naming convention has acted as a catalyst by providing a way to link not only documents, but devices to each other throughout the world.. All that was required was that they be connected to the Internet and have an address, an IP address.
This was possible because the Infrastructure was already in place. And, most important, the Internet did not presume any particular type of usage. The central point is that the Internet itself acts as only a mechanism for connecting end points. It does not decide the policies for the applications.
This gives application developers the freedom and power to create new applications, which can take advantage of this global medium without having to bear the burden of its implementation.
Just as important, it frees the telecommunications network providers to innovate. Since the data traffic is asynchronous and is very tolerant of delays and packet losses, there is a major economic advantage over traditional telephony which is burden by the many demands of a network that is very highly tuned for voice traffic. The change in economics is so great that the Internet is essentially free. A small fee (relative to the costs of traditional telephony) does indeed pay for the Internet.
The chaotic nature of the Internet is a prime contributor to its robustness. Whereas traditional telephony requires a very rigid and thus brittle network, the Internet is able to recover from problems.
The economic advantages of asynchronous data traffic are no longer limited to just low performance data. In the last few years, the distinction between telephony and data communications has been called into question. The concept of isochronous is not fundamental. It is simply a measure of the Quality of Service on the data network. Rather than a sharp distinction between voice and data, there is just the issue of bounding the latency. For streaming data one can tolerate a large buffer to jitter in packet arrival. For interactive traffic the constraints are tighter. For small IP networks, voice can work well. For the larger Internet, we will continue to rely on bridging to the PSTN (Public Switched Telephone Network) until there is sufficient quality in the backbone.
The economics of the Internet make the connected product the norm, not an expensive and specially designed item. Traditionally, products that communicate with each other have required special protocols (or "stacks") that took advantage of the special properties of each implementation. The economics of the Internet strongly shifts the advantage to the use of common protocols. These advantages include:
- The ability to share the common medium.
- The ability to easily communicate world wide.
- The billions of dollars being invested in the design of the global infrastructure
- The Internet is the focus of almost all new protocol design.
- The economy of scale for interfaces being implemented in the millions and, soon, the billions.
IEEE-1394 provides a good comparison. It has received much attention as a means for creating entertainment networks. Are the advantages of IEEE-1394 sufficient to outweigh the problems of creating a new infrastructure. IEEE-1394 requires new. The costs of supporting a very tight timing constraint (such as 125µs) are now greater than the cost of a small amount of local buffering. The real cost of IEEE-1394 is in the inability to take advantage of the protocol design and infrastructure inherent in the IP protocols. While IEEE-1394 can be an IP medium, any implementation that takes advantage of the unique properties of 1394 is limited to its implementation constraints and lack of the intellectual energy that is going into the Internet.
This doesn’t mean that wires should carry only IP protocols or that all devices must be natively IP based. Simply that they must be designed as part of the larger infrastructure. If they can’t be IP natively, they can be represented by proxy service. But native IP devices have the advantage in having control over the participation in the Internet.
Designing Products with the Internet
The focus of product design is on making the products perform their functions as well as they can. The Internet is a means of making the products better. These are not "Internet Appliances". They are simply appliances that are not isolated.
There are a number of simple design points and opportunities:
Communications is should be IP-based. In its simplest form. An IP packet consists of a destination address and data. The UDP (User Datagram Protocol) allows one to simply throw a packet into the Internet. It might or might not arrive at the other end. But even with seemingly reliable protocols, products must be prepared for recovering. Even on 1394. So it is no more difficult to use UDP. There is also the option of using TCP (Transmission Control Protocol) which does provide reliable end to end delivery.
Protocols are designed to IP, not the underlying medium. This allows them to take advantage of the larger infrastructure including the routing capabilities. It is OK to require a high QOS (low latency, very reliable delivery) for some services. They will benefit from the larger scale as the capabilities of the network improve.
Take advantage of the facilities available on the Internet. One can connect to the corporate servers. No longer do you need to ask the user about updates. You can automatically check. You can also sell new capabilities via the network.
Take advantage of the personal computer as a user interface device. You can have specialty applications on the PC or provide a web server in the device. A Web server is a single chip capability.
Use common protocols. There are already common protocols for streaming and device management. This doesn’t mean that one cannot define new protocols. In fact, since the protocols can coexist on the common medium, such definitions and discoveries are part of the vigor associated with the network.
Share data. A DVD CD can be enhanced, for example, with a link to richer information about the movie. When a movie wins an award, the viewer can learn about it even after having purchased the movie.
Just as the Internet allows products to connect to the rest of the world, the products themselves are exposed. As we connect home networks to the Internet, there will be some protection enforced. Longer term, though, products will need to take responsibility for their own integrity. They will be able to participate in authentication protocols that will allow their owners to determine who has what access. The important is not so much in protecting devices. This is already the norm for isolated devices. It is in the ability to selectively grant access and for the owners to access the capabilities remotely. You can readily see who is at your front door (while at the office) and prevent others from doing so.
Examples
What is important about these examples is that they are not new. They are the ones that we have been promising for years but have been unable to deliver because each one must bear the burden of implementing its own infrastructure. Now that we have a common infrastructure and protocols the product ideas become the norm rather than futuristic promises.
VCR Settings
We assume that the VCR is on an IP network so that it can exchange messages with a personal computer and that computer can, in turn, access a program guide on the Internet. The VCR itself need have only the simplest control panel with only the basic functions available. Additional functions become available using the Web form (HTML) in the VCR. It can be viewed on a TV with a browser or on the PC. The PC can add additional capabilities by managing a schedule of programs and coordinating it with the program guide available on the Internet. The basic functions can be provided using a program from the VCR manufacturer or the program guide provider. But third parties can add other capabilities and thus value to the VCR.
Reinventing Telephony
The current voice telephone isn’t integrated into consumer electronics devices. Attempts to add videophones, caller-ID on the TV and other capabilities have largely failed. None of them justify the cost and effort on the consumers’ part to take advantage of them. With the Internet, however, the infrastructure is already in place. The video stream from the camera used to record can simply be redirected over the net and create a casual video feed. The answering machine can readily take messages and forward selectively based on who is being called and add the messages to the same message store as email. Again, not a new idea, but one that has been hobbled by the lack of an effective delivery vehicle. With the devices networked, one can broadcast a "please be quiet" message to other devices in the room without any direct connection among them – just a common IP connection to the home network.
This same notification can allow one to be told that the wash is done or show the picture of the person at the front door on a convenient display surface.
Well, enough for a brief summary.
The Future of the Internet
We are still discovering the Internet. Most important for the future is that the Internet has already become vital infrastructure. We all have too much of a stake to let it fail. It is the way the corporate and governments do business. The economics are so compelling that alternate special networks cannot be justified.
Of course, as this becomes the common medium, security and integrity are necessary. And ubiquitous security (IPSEC protocols) is coming to the fore. Other protocols will support authentication and trust relationships to make it easy to specify who has what access. You can share your photo albums with your friends and family and no one else. You can specify who can open the garage door. Unlike a metal key, you simply add them to the access list and they can use their standard IP-based door opener with no additional devices.
One big hurdle is the quality of service in the backbone network. Currently the behavior makes high performance interactive streaming problematic. Major investments in the backbone help here as do new protocols. Multicast protocols, for example, will greatly reduce the cost of reaching millions of users simultaneously.
The personal computer is also changing in response to the Internet. The capabilities of the personal computer are becoming available throughout the house and beyond. Many users are buying multiple PCs. They can communicate with each other and thus one has immediate access to the computers wherever one is. And much of this access is Web-based and there is the option of simple devices for limited Web access.
The Internet is Now
While there is much room for improving the Internet, we must not forget that the Internet is already the common medium. And that using Internet Protocols (IP) will take advantage of the current and future connectivity.
It will also allow one to take advantage of the personal computers that many customers already have. A VCR schedule can be set using a personal computer with a richly graphical display and with access to the full program schedule available over the Internet. The PC can also manage multiple recordings on a single digital tape. It can even scan for missing episodes when recording a series of programs.
Devices that do not take advantage of the Internet will suffer by comparison. They will be missing capabilities that the customer expects. It is only recently that VCR clocks get their time from the time transmitted in the subcarrier for many TV stations. Such a small step took a long time to arrive. The enhancements available via IP are much more dramatic. We shouldn’t have to wait for them.
A Simple Framework for Cooperating Devices
This is a proposal for an approach for defining interfaces for devices. It builds upon a proposed Web protocol (XML, see http://www.w3.org/XML). It only defines the framework for the interfaces and allows industry groups or individuals to define common interfaces.
Unlike more traditional protocols involving long design efforts, the protocols on the Internet begin with a simple design and then evolve over time to support additional capabilities. The goal of this simple framework is to provide a simple set of rules for creating cooperating products without waiting for a full design.
The goal of this framework is to allow product designers full freedom to create protocols while providing a mechanism for cooperating. We create a marketplace for interfaces by defining just enough to allow some cooperating without limiting the individual efforts. There is a strong reward for agreeing on common definitions because the devices that agree benefit from the synergy with other such devices.
The most important characteristic such a framework is that it be resilient. We expect errors in both design and in the operation of devices. This means that the two devices must be tolerant of any errors and misbehavior in other devices.
Erring on the side of simplicity, this framework ignores the more complex issues. It just provides a stateless mechanism for discovering the characteristics of other devices and operating them. It must be supplemented by other protocols for specific operations. For example, video-streaming protocols are beyond this framework.
The basic framework is built upon interfaces. An interface is a collection of tags grouped together in a namespace. There would usually be agreement over common interfaces definitions such as that for a printer though this is not a requirement. A device would have multiple interfaces. A printer can also have a "physical location" interface that tells where it is and a picture of itself. Another interface would support security capabilities allowing an administrator (machine or person) to set policies.
XML
The W3 consortium is defining an extension to HTML called XML. While there are many goals of this extension I’m focusing on one aspect – the ability to define sets of tags for data values on an HTML page.
We can define tags for a given interface. For example, a printer can publish its characteristics as a set of XML "printer" definitions. Since these namespaces are defined by unique tags, anyone can define a printer set. In practice there will be a preferred set of tags (the namespace) for printer attributes. Thus we can get agreement over common cases while allowing full freedom to define additional interface
Rendezvous
How do two devices find each other? In a very small network, they can broadcast their names and a list of the interfaces they support. Machines entering this network could request such a broadcast. For larger networks and for routing to the Internet as a whole we’ll need directory services. At a low level, a dynamic DNS (Domain Name Service) can maintain a mapping from a stable unique or descriptive name to the current IP address for the device.
At a higher level, a full directory service can allow for finding devices according to characteristics and administrator policies.
Conclusion
The Internet is the new infrastructure. It requires that we rethink the nature of devices. They no longer need be isolated. The cost of tapping into the global network is very low and the new capabilities are compelling. The designs need not be limited by what the device can do alone.
We need to learn how to create such devices. What capabilities should be available locally and which can be shifted to a more sophisticated device such as a personal computer. How does a device maintain a relationship to the manufacturer?
In order to take advantage of these opportunities we need simple and robust protocols. Unlike an isolated device where one manufacturer can take full responsibility for the devices’ function, the new device can be affected by the misbehavior of other devices and must be resilient even when other devices fail.
A device that cannot cooperate with other devices will be like the typewriter – a story for one’s grandchildren. Like the days we used to multiply numbers by using a slide rule.
References
In keeping with the spirit of this paper, the references are URLs:
| XML | http://www.w3.org/xml Note that this is a complex proposal and I’m only concerned about a small portion of the specification. |
| IPSEC | http://www.ietf.org/html.charters/ipsec-charter.html |
| IPV6 | http://www.ietf.org/html.charters/ipngwg-charter.html |
| IEEE‑1394 | http://www.firewire.org/ |
The slides for the talk
Updated in 2023 to include the talk itself. In 1997 it seemed cool to have garish slides. Times have changed
Acknowledgments
I would also like to thank Dr. David P Reed for his helpful insights on Internet Protocol issues as well as his participation in their initial design.
